Worked on stabilizing and enhancing CI/CD pipelines across NVIDIA-NeMo/Megatron-Bridge, NVIDIA-NeMo/Automodel, and related repositories, focusing on improving reliability, test infrastructure, and production readiness for large-scale machine learning workflows. Addressed silent failures in CI by refining failure reporting and enforcing consistent repository metadata, while optimizing Docker environments and introducing conditional test gating for GB200 hardware. Upgraded core dependencies and synchronized test suites to maintain compatibility, using Python, Bash, and GitHub Actions. Enhanced test scalability with GPU sharding and centralized NCCL environment management via Pytest, resulting in faster triage, more reliable releases, and streamlined development for distributed systems projects.